Вы находитесь на странице: 1из 12

1

Chapter 6: Isoparametric elements

• Same shape functions are used to interpolate nodal


coordinates and displacements
• Shape functions are defined for an idealized mapped
element (e.g. square for any quadrilateral element)
• Advantages include more flexible shapes and
compatibility
• We pay the price in complexity and require numerical
integration to calculate stiffness matrices and equivalent
loads

R.T. Haftka EML5526 Finite Element Analysis University of Florida


2

Bar element example


• Three-node bar example used only for illustration

• Quadratic variation of both coordinate and displacement


in terms of ideal element coordinate
a4  a4 
   
x = 1 ξ ξ2  a5  and u = 1 ξ ξ2  a5 
a  a 
 3  6
−1
 x1  1 −1 1  a1  1 −1 1  x1 
    a   
x
 2 = 1 0 0   2 hence x = 1 ξ ξ 2  1 0 0  x2 
 x  1 1 1 a   
 3    3 1 1 1  x3 
R.T. Haftka EML5526 Finite Element Analysis University of Florida
3

Interpolation functions

T
x =  N   x1 x2 x 3 
T
u =  N   u1 u2 u 3 

1 1 2 
 N  =  ( −ξ + ξ ) 1 − ξ ( ξ + ξ )
2 2

2 2 

R.T. Haftka EML5526 Finite Element Analysis University of Florida


4
Strains and stiffness matrix
• Because of coordinate transformation taking derivatives
is more complicated
u1 
du  d   d dξ d
ε x = =  N  u2  where =
dx  dx   dx dx dξ
• Jacobian u3 
 x1   x1 
dx d   1 1  
J = = N  x2  =  ( −1+ 2ξ ) −2ξ (1+ 2ξ )  x2 
dξ dξ x   2 2  
 3 x3 
• Hence strain matrix
1 d 1 1 1 
B = N  =  ( −1+ 2ξ ) −2ξ (1+ 2ξ )
J dξ J 2 2 
• And stiffness matrix
L 1

k  = ∫ B  E B  Adx = ∫ B  AE B  Jdξ
T T

0 −1
R.T. Haftka EML5526 Finite Element Analysis University of Florida
5

6.2: Bilinear quadrilateral


• We were limited to rectangles because of compatibility

• Now both displacement and coordinates are bi-linear


functions of ξ and η
• How does that preserve compatibility?

R.T. Haftka EML5526 Finite Element Analysis University of Florida


6

Interpolation
• Mapping
 x   ∑ N i xi  u   ∑ N i u i 
 =  = [N ]{c} &  =  = [N ]{d }
 y  ∑ N i yi   v   ∑ N i vi 

{c} = x1 y1 x2 y2 x3 y3 x4 y4 
T

{d } = u1 v1 u2 v2 u3 v3 u4 v4 
T

 N1 0 N2 0 N3 0 N4 0 
[N ] = 
0 N1 0 N2 0 N3 0 N 4 
• Interpolation functions
1 1
N1 = (1 − ξ )(1 − η ) N 2 = (1 + ξ )(1 − η )
4 4
1 1
N 3 = (1 + ξ )(1 + η ) N 4 = (1 − ξ )(1 + η )
4 4
R.T. Haftka EML5526 Finite Element Analysis University of Florida
7

Derivatives
• Chain rule of differentiation
∂φ ∂φ ∂x ∂φ ∂y
= +
∂ξ ∂x ∂ξ ∂y ∂ξ φ , ξ  φ , x 
or   = [J ] 
∂φ ∂φ ∂x ∂φ ∂y φ
 η, φ
 y,
= +
∂η ∂x ∂η ∂y ∂η

• Jacobian

 x ,ξ y ,ξ   ∑ N i ,ξ xi ∑ N ξ y 
[J ] =   = i, i

 x ,η y ,η   ∑ N i ,η xi ∑N ηy  i, i

R.T. Haftka EML5526 Finite Element Analysis University of Florida


8

Derivatives-completed

• Using shape functins


 x1 y1 
 
1 − (1−η) (1−η) (1+η) − (1+η)x2 y2   J11 J12 
[J ] =   = 
4 − (1 − ξ ) − (1 + ξ ) (1 + ξ ) (1 − ξ )  x y 
 3 3  21 22 J J
 
x y
 4 4
• Then
φ , x  Γ11 Γ12 φ ,ξ  1  J 22 − J12 
  where [Γ] = [J ] = 
−1
 =  
φ , Γ Γ
 y   21 22  η  φ , J − J
 21 J11 
[Γ ]

R.T. Haftka EML5526 Finite Element Analysis University of Florida


9

Strain and stiffness matrices

• Strains u, x 
 ε x  1 0 0 0  
    u, y 
{ε } =  ε y  = 0 0 0 1 
γ  0 1 1 0 v, x 
 xy    v, 
 y
• Where
 u , x   Γ11 Γ12 0 0   u ,ξ 
u ,    
 y   Γ 21 Γ 22 0 0   u , η 
  =  
 v,x   0 0 Γ11 
Γ12  v , ξ 
 v , y   0 0 Γ 21

Γ 22   v , η 

R.T. Haftka EML5526 Finite Element Analysis University of Florida


10

Laboring to obtain B matrix


• Finally expressed in terms of nodal displacements
u ,ξ   N1,ξ 0 N 2,ξ 0 N 3,ξ 0 N 4,ξ 0 
u ,   N 0 N 2,η 0 N 3,η 0 N 4,η 0 
 η
  =  { d}
1,η

 v ,ξ   0 N1,ξ 0 N 2,ξ 0 N 3,ξ 0 N 4,ξ  1*8


 v ,η   0 
N1,η 0 N 2,η 0 N 3,η 0 N 4,η 

• So B matrix obtained by multiplying these 3x4, 4x4 and


4x8 matrices

• Stiffness matrix
1 1
[k ] = ∫∫ [B ] [E ][E ]t dx dy [ ] [E ][B ]t J d ξ d η
T
∫∫
T
= B
8 *8 8*3 3* 3 3*8 8*3 3* 3 3*8
−1 −1

R.T. Haftka EML5526 Finite Element Analysis University of Florida


11

6.3: Numerical quadrature


x2 1
I = ∫ f dx becomes I = ∫ φ dξ
x1 −1

• With the transformation:


1 1
x = (1 − ξ )x1 + (1 + ξ )x2
2 2
• Numerical quadrature
1
I = ∫ φ d ξ ≅ W1φ1 + W 2φ 2 + .......... . + W nφ n
−1

R.T. Haftka EML5526 Finite Element Analysis University of Florida


12

Gauss quadrature

• How to select optimum points and weights

R.T. Haftka EML5526 Finite Element Analysis University of Florida